SF TimeStats: fix a hazard issue
The setAcquireFence and setDesiredTime could happen in between the
splited setLayerName and setPostTime, where setLayerName inserts an
entry but setPostTime actually initializes the structure and waitData
position. This change combine setLayerName and setPostTime again to
eliminate the hazard.
Test: build
Bug: b/118770127
Change-Id: Ia1961cb688bbabf23d88305750e69c9fa41da8dd
diff --git a/services/surfaceflinger/BufferStateLayer.cpp b/services/surfaceflinger/BufferStateLayer.cpp
index 5df8ade..73098bf 100644
--- a/services/surfaceflinger/BufferStateLayer.cpp
+++ b/services/surfaceflinger/BufferStateLayer.cpp
@@ -470,8 +470,7 @@
}
// TODO(marissaw): properly support mTimeStats
- mTimeStats.setLayerName(layerID, getName().c_str());
- mTimeStats.setPostTime(layerID, getFrameNumber(), latchTime);
+ mTimeStats.setPostTime(layerID, getFrameNumber(), getName().c_str(), latchTime);
mTimeStats.setAcquireFence(layerID, getFrameNumber(), getCurrentFenceTime());
mTimeStats.setLatchTime(layerID, getFrameNumber(), latchTime);